Are we having a good time?

The fundamental time interval is the SI second, a part of the International System of Units (SI). Every other measure of elapsed time is only an accumulation of seconds. Before the SI second was benchmarked to the fundamental properties of matter, it was based on a consensus average of the rate at which the Earth rotates. However, the Moon steals angular momentum from the Earth, and the Earth's rotation rate is generally decreasing. Thus the old ‘standard’ second, 1/86,400 of a day, was not constant. It was painful to calibrate transmission frequencies based on a variable second. The SI second, formalized in 1967, overcame this problem.¹ But astronomers, who need to find objects in inertial space from this moving and rotating planet, must keep track of the Earth's wobble and rotation. This task was straightforward when time was measured in fractions of Earth rotation. Disconnecting the second from Earth rotation would burden them.

Coordinated Universal Time (UTC) was a compromise, measuring time in very stable standard seconds while including occasional corrections to synchronize time with Earth rotation. Beneath this civil adjustment, the difference between time in SI seconds and time in Earth rotation (called UT1) is also broadcast within the UTC scheme for those who require greatest precision. Scientists at the International Earth Rotation and Reference System Service in Paris keep track of the accumulating discrepancy between UTC and UT1 and declare when a second should be added or subtracted. The goal is to keep the absolute discrepancy less than the benchmark value of 0.9 SI seconds. Deep inside much system software there is a counter that waits for a threshold value of the accumulated discrepancy and acts to accept a leap second when the magic value approaches. The UTC specification recommends that leap seconds be implemented either on the last day of June or the last day of December. In practice, the discrepancy is somewhat less than 0.9s when a leap second is applied, but it can never be greater than 0.9s. Whatever that value, when a leap second is added the value of the discrepancy is adjusted down appropriately.

More than a decade ago, prominent members of the International Telecommunication Union (ITU) decided that correlation with Earth rotation was at best an inconvenience. They felt that confusion about how to accommodate leap seconds impaired important tasks. This confusion could happen because there is no normative authority for defining civil time and no uniform guidance for responding to warnings of imminent leap seconds or accommodating them. This situation pertains even though nations have declared UTC, which has no official definition or legal control, as their statutory time scale. The United States, for example, has a statute mandating UTC to be the legal civil time scale, leaving it up to the US Naval Observatory and the Department of Commerce to define what UTC is. The ITU has no enforcement authority. The current specification for UTC is Recommendation ITU-R TF.460-6.² A recommendation is not a normative decree, and it can (and might) be changed without legal due process.

Doing away with leap seconds, however, is fraught with problems.³ Space system operators must be able to correlate precise civil time with Earth rotation. The Earth moves about one angular degree around the Sun each day. If one sees a star or other object in the heavens at a precise time today and in a precise direction, it will not be in the same direction from the same place at the same civil time tomorrow.⁴ Losing the only widespread and easily accessible correlation with Earth rotation would have a serious impact. If synchronization with Earth rotation were deprecated, the difference between UTC and UT1 would grow beyond 0.9s, violating constraints in widespread software, leading to imponderable emergent behavior in software that previously was reliable and trustworthy.

Some people claim that the leap second makes UTC discontinuous. On the contrary, even with leap seconds, time evolves smoothly, second by second by second, just as the accumulation of seconds in days and months continues though the number of seconds we assign to each month might be irregular. Digital clocks have no problem displaying an extra second or skipping one. Analog timepieces might resemble the one in Figure 1.

Figure 1. A leap-second stopwatch. Digital time pieces can easily count continuously through leap seconds. Analog instruments cannot, but they are not sufficiently precise for it to matter.

Several excellent texts bring perspective.^1,5 The current conundrum at the level of small fractions of a second is just the latest among many time-keeping inconveniences brought on by the profound lack of cooperation among the Earth, the Sun, and the rest of the universe. Each element goes its own way without pity on earthlings who must measure time in inconsistent Earth rotation but work synchronized with seasons and important milestones in the heavens. One cannot count days one by one and arrive at exactly one complete circuit of the Earth around the Sun. Nor can one divide the useful day from sunrise to sunset in a consistent, convenient, and enduring way. The current leap second issue follows numerous but far from universal calendrical manipulations. The most notable was Julius Caesar's one-in-four leap years, which turned out not to be quite correct, leading Pope Gregory to leave out 11 days, 1500 years later.

No approach will meet everyone's needs or eliminate all inconvenience. No one knows the extent of the impact if UTC were changed so significantly, eliminating leap seconds. On the other hand, no one knows the extent of difficulties with the current scheme. The ITU's Radio Assembly and World Radio Conference earlier this year acknowledged as much by deferring a decision for at least three years so that issues could be quantified and mitigations conceived for both possible outcomes.⁶

The treacherous pathway through time will continue forever. The next leap second, the first since January 2009, is scheduled to be added at 23h 59m 60s on 30 June. This event is an opportunity to monitor leap-second implementation and capture issues. See our recent papers for more on this matter.^{3, 4}